1. Field of the Invention
The present invention relates to a technology of a resource node assignment in a distributed computers environment.
2. Description of the Related Art
Conventionally, when access from users concentrates on one service, or a service requiring a large quantity of calculation jobs is performed, for the load sharing, a distributed computer environment is used where a large number of computers are connected via networks, for example, a grid environment in which different types of computers are intermingled.
In this case, the optimal computer (hereafter “resource node”) suitable for the purpose of one service must be selected automatically and quickly among a large number of computers to be assigned to the service. Especially, in the grid environment in which different types of computers are intermingled, the performance of the whole service depends strongly on the assigned resource node.
For this reason, to improve the performance of the whole service, a resource node depending on each service is assigned based on, for example, information on the performance of each resource node and a use environment.
Specifically, for example, a service administrator requests a resource required for the service from a resource broker that performs the brokering of resource nodes used among a plurality of services. A resource broker searches a resource node that satisfies this request from a resource node group, and assigns a resource node used for service based on the search results (for example, Japanese Patent Laid-open Publication No. 2002-024192, Published Japanese Translation of PCT application No. 2004-521411, and Japanese Patent Laid-open Publication No. 2006-031358).
However, in the conventional techniques of the patent documents 1 to 3 mentioned above, the initiation instruction of an application required for service is performed to each resource node after being assigned as the resource node used for the service. Namely, the assignment of a resource node used for one service is performed, while the initiation of the application in each assigned resource node is not guaranteed.
For this reason, there has been a problem that the service could not be provided because of a resource node required for performing service not being prepared, when, in each resource node, a host-down has occurred just before the initiation of an application, an initiation instruction could not be received due to a temporary communication trouble, or the like.
Moreover, at the time of selecting the resource node actually used for the service among resource nodes capable of providing the service, the selection criterion is uniquely decided by the resource broker.
However, since the selection criteria differ for every service in many cases, the selection criteria for every service must be understood beforehand and a resource broker based on them must be developed. For this reason, there has been a problem that when services to be performed increase or distributed computer environment is used for other services, the addition and change of selection criteria cannot be flexibly performed, resulting in poor versatility.
To solve the aforementioned problems in the conventional techniques, the present invention aims at providing a resource node determination program realizing the smooth provision of each service, a resource node control program, a recording medium, a resource node determination apparatus, a resource node, a resource node determination method, and a resource node control method.